GD-Wheel

  • Li C
  • Cox A
  • 14

    Readers

    Mendeley users who have this article in their library.
  • N/A

    Citations

    Citations of this article.

Abstract

Memory-based key-value stores, such as Memcached and Redis, are often used to speed up web applications. Specifi-cally, they are used to cache the results of computations, such as database queries and dynamically generated web pages, so that a future request to the web application may not have to repeat the same computation. Currently, when memory-based key-value stores reach their capacity limits, they use replacement policies, like LRU and random, that are oblivi-ous to differences among the cached results in their recom-putation costs. However, this paper shows that if the costs of recomputing cached results vary significantly, as in the RUBiS and TPC-W benchmarks, then a cost-aware replace-ment policy will not only reduce the web application's total recomputation cost but also reduce its average response time. To this end, this paper introduces GD-Wheel, which is an amortized constant-time implementation of the GreedyDual replacement algorithm that supports a limited range of costs. In effect, GD-Wheel integrates recency of access and cost of recomputation in an efficient manner. Moreover, this paper describes an implementation of GD-Wheel in Memcached, including the modifications to Memcached's interface so that web applications can include cost information with each key-value pair, and a new cost-aware slab rebalancing policy for Memcached's slab-based memory allocator. An evalua-tion of this implementation using the Yahoo! Cloud Serving Benchmark shows that GD-Wheel, when compared to LRU, reduces the total recomputation cost by as much as 90%. Moreover, GD-Wheel reduces the web application's average

Author-supplied keywords

  • application-level
  • replacement policy

Get free article suggestions today

Mendeley saves you time finding and organizing research

Sign up here
Already have an account ?Sign in

Find this document

Authors

  • Conglong Li

  • Alan L. Cox

Cite this document

Choose a citation style from the tabs below

Save time finding and organizing research with Mendeley

Sign up for free